MIT CSAIL's Alex Zhang clarifies that his recursive language model research only connects to two Anthropic agent systems
These are Anthropic's Scaling Managed Agents and dynamic workflows.
@xeophon not the training mines...
Florian Brand@xeophon@a1zhang stop wasting your time with ppl unable to read and come into the training mines instead
@a1zhang ignore the haters and keep doing good work 👍
alex zhang@a1zhangLast thing and then I'll shut up because people on this website genuinely have trouble 1) reading and 2) remembering anything: I have referenced something else not openly called an RLM no more than 2 times as being RLM-like. Both were Anthropic releases. The first was "Scaling Managed Agents" post which explicitly links to the RLM paper as inspiration (and no one has issue with), and the second was the most recent "dynamic workflows" release. That's literally it. And all I said was "Every day we move closer to the RLM". I haven't deleted any tweets, you can check this yourself. People are acting like I've egregiously done this over the past year and have hyped up this paper to oblivion because you see your favorite illiterate influencer shitpost about it. Genuinely fuck off, you're remembering all the bullshit AI influencers that were spouting nonsense about "RLMs solved long-context" or "RLMs were the first to do sub-agents". @yacinelearning remembers all of this and even made a good YT video about it, and while I am sorry that your TLs were polluted with it, it has nothing to do with me or the paper (half the time they butchered the name anyways). If you're sick of that hype, fine, don't take it up with me, I don't condone it either. Since the release, every time I go on this website and see an RLM-specific paper or tweet I like, I retweet it. That's it. Stop acting like every release I claim is a new RLM feature. IT'S LITERALLY JUST THIS ONE. That's all. If you have more problems, take it up in DM. I'll even call you if you're that desperate. This is so fucking stupid man
@a1zhang stop wasting your time with ppl unable to read and come into the training mines instead
alex zhang@a1zhangLast thing and then I'll shut up because people on this website genuinely have trouble 1) reading and 2) remembering anything: I have referenced something else not openly called an RLM no more than 2 times as being RLM-like. Both were Anthropic releases. The first was "Scaling Managed Agents" post which explicitly links to the RLM paper as inspiration (and no one has issue with), and the second was the most recent "dynamic workflows" release. That's literally it. And all I said was "Every day we move closer to the RLM". I haven't deleted any tweets, you can check this yourself. People are acting like I've egregiously done this over the past year and have hyped up this paper to oblivion because you see your favorite illiterate influencer shitpost about it. Genuinely fuck off, you're remembering all the bullshit AI influencers that were spouting nonsense about "RLMs solved long-context" or "RLMs were the first to do sub-agents". @yacinelearning remembers all of this and even made a good YT video about it, and while I am sorry that your TLs were polluted with it, it has nothing to do with me or the paper (half the time they butchered the name anyways). If you're sick of that hype, fine, don't take it up with me, I don't condone it either. Since the release, every time I go on this website and see an RLM-specific paper or tweet I like, I retweet it. That's it. Stop acting like every release I claim is a new RLM feature. IT'S LITERALLY JUST THIS ONE. That's all. If you have more problems, take it up in DM. I'll even call you if you're that desperate. This is so fucking stupid man
